On Fri, 18 Jan 2019 at 21:49, Kevin Kenny <kevin.b.ke...@gmail.com> wrote: > > We deal with indefinite objects more often than some people are > comfortable with. (I've mentioned previously that my state has such > things as county lines that are in part unsurveyed!) > > Rather than a new relation type, I think it would be simpler to tag > the indefinite part of the boundary of whatever area feature with a > key like "indefinite=yes". An indefinite boundary will normally have > no reason to have tags of its own other than this one - because it > would need to be a 'real' feature in order to have most of them be > meaningful. It would ordinarily be there only to close a multipolygon > topologically, and the tags of the multipolygon of which it's an inner > or outer way would ordinarily be the only other information pertaining > to it. > > If we try to fix "maximal" and "minimal" area, we'll simply run into > more haggling- because the maximum and minimum do not have bright-line > definitions, any more than the indefinite line does. We'll have > interminable arguments over what land might and might not be > considered part of a peninsula. I'd like to nip that in the bud by > simply declaring that any choice is arbitrary, and that the drawing of > an arbitrary boundary of an area feature should be informed in part by > what the locals think. Is Wareham, Massachusetts on Cape Cod? I have > no idea, but I bet that the locals have a rough consensus - and if > they don't, that they'd at least be unsurprised if a mapper were to > choose the Cape Cod Canal or the Plymouth County line as the cutoff > with an 'indefinite' indication. > > Simply having the tagging allow for an 'indefinite line', I think, > could be a near-universal solution to the fact that bays, peninsulas, > channels, isthmuses, lakes with broad inlets/outlets, rivers with > broad mouths, administrative regions with unsurveyed boundaries, > mountain ranges, etc. all are area features that have a distinct > shape, except for the fact that part of their margin may be > indefinite. > > Try as we might to make them go away, there are objects, observable > and named in the real world, that are areas, part of whose boundaries > are indefinite. Saying that such things can be only point features is > shortsighted.
The only imperfection of indefinite=yes tagged only on the way that connects the peninsula to the mainland is that this doesn't make it explicit that part of the coastline – while not fuzzy by itself – might or might not be part of the peninsula. (For example, part of the coastline between – or even beyond – these two point [^1][^2] might or might not belong the Presqu'île de Crozon [^3].) Tagging the relation member roles for example outer:indefinite were less imperfect, but still imperfect – like my previous idea with the minimal and maximal area. Besides, it were incompatible with the current type=multipolygon specification. Another solution would be to admit that peninsulas – as well as bays, channels etc. – inherently have fuzzy borders and that therefore tagging the fuzzy borders differently is unnecessary. [^1]: <https://www.openstreetmap.org/node/2007095790#map=13/48.2706/-4.2582> [^2]: <https://www.openstreetmap.org/node/534716211#map=13/48.2706/-4.2582> [^3]: <https://www.openstreetmap.org/relation/9238102> Regards Markus _______________________________________________ Tagging mailing list Tagging@openstreetmap.org https://lists.openstreetmap.org/listinfo/tagging
